Contribute
Register

Z490 & Z590 - Will Z590 ever have macOS Support ?

Yeah, works fine in Windows for me as well..Just black screen in macOS :(
Doesn't seem to work in Manjaro Linux either when using the USB-C cable to the monitor.
Try adding i915.force_probe=4c8a as a Linux grub boot argument. I had to do this on Ubuntu to get the iGPU working.
 
Which kernel version are you using? I think the Rocket Lake iGPU support landed in the later 5.11 kernels and should be there in 5.12. I know also one point the AX210 chip (which I swapped in) wasn’t supported in 5.11 but it is on 5.12.
5.12

Try adding i915.force_probe=4c8a as a Linux grub boot argument. I had to do this on Ubuntu to get the iGPU working.

Will try that. Thanks!
 
It's barely out of the factory, I suppose. :)
I haven't seen it yet at any retailer, but this card seems rather promising due to its use of standard 5-pin and 3-pin Thunderbolt headers. No fancy 14-pin headers or other such monstrosities needed.
 
Sorry I thought it was out already, looks very interesting for sure.
 
Oh, it's finally announced. I've been using the ASUS ThunderboltEX 4 (haven't tried Thunderbolt yet, just USB).
https://forums.macrumors.com/threads/usb-3-x-pcie-cards-for-classic-mac-pro.1501482/

The ASRock Thunderbolt 4 aic is also another alternative with a 5 pin header. Still not available at retailers, though.
They're all coming out of the woodwork (but not in the negative way that idiom implies - there's probably a more appropriate one for this news).

The Asrock is interesting because of the jumpers for "Force Power on" and "Force Reset" and "No RTD3". RTD3 is for PCIe Runtime D3. I suppose they're not necessary if you can do the same thing by connecting pins of the header cable to ground or 3.3V (or whatever the required voltage might be).
 
Followup on my problems with Sabrent Rocket 4 NVMe. The original Rocket lasted about 3 weeks before developing a problem with stalling and spontaneous disconnects. I was in the middle of system tuning and BIOS upgrade from beta 0704 to 0707 when it died. It took several weeks and I obtained a replacement from Sabrent customer support.

The replacement appears to fresh off the production line, and it lasted 3 days before suffering the same fate as the first one.

However all was not well pretty much from start. I ran into trouble trying to make a clone from a working EVO 970 using CCC with the APFS replicator giving up with an error. To add to complexity, Bombich SW has been releasing frequent updates to Carbon Copy Cloner to deal with new Apple sealed system volume and the requirement to use Apple's APFS replicator to make bootable clones. There's a lot of jabbering about bootable M1 Macs that attends these changes so I have lots of things in flux, OpenCore, BIOS, overclock tuning, significant system updates.

After seeing problems with the APFS replicator, went to Linux and tried reading the raw drive with ddrescue. This showed severe read-stalling early in the drive extent. Not knowing what else to do, I zeroed the drive using ddrescue to examine it for defects. This operation stalled at first, like reads, then quickly moved on and when it was done the drive behaved well with high expected performance.

From here I decided to clean install Big Sur 11.3.1 from the EVO and migrate my stuff. I ran into a glitch with accessing the macOS installer dmg that turned out to be related to my RAM. My build has otherwise been running fine. At one time I was playing with nudging memory timings to see how this affected bench-marking results, and I had left the RAM at XMPII + 200Mhz. Following a lead from a forum, I downclocked RAM to XMPII and behold the dmg mounted.

I moved forward with the system installation which proceeded at record speed, like 20 minutes from starting the installer to having all my data migrated! I dig this technology! Along the way I noticed the Rocket 4 getting hot to touch! But it never slowed down. I was running it open-air in slot M.2_3 with no fan. I moved a case fan over to it and ran a contrived sequential read test at the terminal by mixing find with dd to grab big files and dump them at /dev/null as fast as it will go. Doing this and increasing the found-file size, I discovered that Rocket 4 temps would surge well into mid 70s C. Sabrent lists maximum operating temp at 70C.

I removed the Rocket 4 from slot 3 and placed under heat-sink in slot M.2_2.

it just so happened that a day before I was looking at an iFixit user thread about high CPU load after installing an SSD in an older Macbook, and there was the idea that the solution was to replace the internal SATA cable because data was getting all choked-up because of unreliable SATA I/O. It didn't occur to anyone that Spotight gets let off the leash with an SSD, so I wrote up an explanation and submitted it there.

After the system restarted I kicked off indexing using the Spotlight Privacy preferences and watching drive temps with iStat menus. The Rocket under the heatsink went through the roof hitting 87 C in not time, then the system stopped.

I looked more carefully at slot M.2_2 from a thermal point of view and noticed that the wide flat bottom of the W5700 GPU is covering half the heat sink and resting on the heat sink, so I contrived a shim and moved the case fan over to get air going across the top of the heat sink.

I restarted and the indexing resumed with Rocket temps hitting mid-70s.

From here I reached out to Sabrent to explain what I was seeing, but it takes one day per email exchange with them.

In the mean time, as they previously said they want to know what Rocket Control panel says, and I am wondering what formware rev the Rocket is running, I spend an entire day getting Windows 10 installed on an old spinner, which is just a horrible experience with the installer doing crazy things that I won't get into.

And this spinner install ends up so slow and painful that I spend another day getting an SATA SSD with Ubuntu on it to dual boot Windows. Along the way I accidently wrecked my Ubuntu and had to reinstall, then I met the challenge of dd-ing a working windows installation over to a partition on the SSD and getting it to harmonize. Windows is pretty well behaved about multi-boot these days, it respects existing contents of the ESP (EFI service part). But damn Linux keeps trumping the NVMe ESP, ... Why? ... because it's there, even though I have explicitly directed it to install on anther drive and taken care to tell it to use that drive for boot. YARRRG. I submitted a bug report to Canonical about this when it happened to me a couple of weeks before as I was trying to get OEM kernel runnning to test this board's iGPU.

As an aside, I can say that Ubuntu and Windows handle iGPU+dGPU multi-monitor on this build completely without so much as a glitch.

So now that I can run Rocket Control panel, I find out that my firmware is the latest, but it does pretty much nothing else. Sabrent does not offer a "secure erase" that resets the drive. Rocket Control panel is very little more than a list of buttons to access Windows built-in utilities., such as Device Manager and Computer Management and print some smart info.

I'm feeling pretty good overall at this point because I've got 3 systems booting, Mac is 11.3.1 is up on new Rocket. There's a temp concern which I'm engaged with Sabrent support about.

Also, NVMefix runs for me on 11.3.1 and I discover by accident that sleep works, although it takes two key presses to wake, one to wake the board and another to wake the display. I disable system sleep because I don't want to worry about this for now.

I end another day thinking I'm good-to-go and leave the system idling that night. The next day I think I'm going to get back to work and the screen doesn't wake up when I touch the keyboard. But the board status LED is normal running A0. I force reset and no macOS comes up, it wants to go to Ubuntu. I get into BIOS and the NVMe settings show no device. The replacement Rocket 4 has died!

From here I pull it out and replace the EVO to get going again, and connect the Rocket via the Sabrent external enclosure. It mounts like everything is fine, but after a few minutes it spontaneously disconnects. From here it just gets worse. I take it back to Linux and Windows and it shows up in various guises but I/O is completely unreliable. Linux Gparted shows an ESP and APFS. When I get back to macOS on EVO, the drive appears unformatted. I leave it attached to the type-C port for a while to work on something else and a few minutes later when I pick it up, the Sabrent aluminum case is so hot I think I'm being burned and reflexively drop it. That enclosure was hella hot!

Now messages between me and Sabrent support are crossing like ships in the night, with them asking me to run "chkdsk /r" and me saying "But I'm not running Windows on this drive" and "BTW it's dead!"

The same day I got the replacement Rocket from Sabrent, I had the presence of mind to order another drive, and chose a Western Digital SN750 2 TB. It arrived today and I quickly and easily clean-installed 11.3.1 and migrated all my settings from the EVO at record speed. I ran the same stress tests as the Rocket and it's peak temp open-air is 60s C, and under a heatsink it peaked at 59 C, and temp went down to 50 C when Spotlight indexing. These conditions caused the Rocket 4 to blow.

I'm using the WD SN750 to type this report and while it's still very early, I have every reason to expect that this drive, like the EVO 970, will run fine. (fingers crossed)

I updated my build-page from earlier on this thread with some additional details about the Rocket 4 and the Asus M.2 slots and specs.

* * *

THERMALS DRILL-DOWN

Here are some numbers on Sabrent Rocket 4 (replacement drive) thermals, open-air and under slot M.2_2 heat-sink.

Sabrent-Rocket-Over-Temp.png


The spikes at left are a series of sequential Mac file accesses done using find with dd if={} of=/dev/null bs=1m. I ran several passes to watch how temps trended. A cooler is needed, but the drive didn't appear to slow down.

I shut-down and moved the Rocket into slot M.2_2 under its heat-sink and reset Spotlight indexing: BOOM.

Why did this spike happen?! The heatsink was too hot to touch.

The Rocket 4 survived after the crash (it's just the Rocket that crashed, rest of the system was still alive). It quickly cooled and at reboot it appeared to work normally. But the next day the drive was dead, just like the first one.

Here's the Western Digital SN750 under slot M.2_3 heat-sink (an open-air peak of 67 C encountered during the frst phase of macOS installation is not shown):

WD-SN750-File-Copy-and-Indexing-Temp.png


Here's a look at the Rocket 4 in Windows Sabrent Rocket Control Panel after the drive has completely failed.

According to the Control Panel what is the state of the drive?

BD20095C-04DE-40FE-9995-F2FC4A0A84DA.jpeg


The drives behavior is completely erratic at this point: Rocket Control Panel Tools and SMART report may cause a hang. Windows Disk Management Initialize dies with I/O error In Linux, the drive can sort of be written to, Gparted sees a GPT but data reads may hang. In macOS it appears uninitialized and cannot be formatted, it disconnects.

As with the first drive the collapse was sudden and occurred while the drive was pretty much idle. As with the first drive ALL DATA WAS LOST. But other than this, it appears to be fine?

If you can barely read the above screen-shot, that's just Sabrent's idea of a modern UI design. Over on the right, in the warm copper-plate shadows, I've highlighted some stats: SMART power-cycle count of 282 and unsafe-shudowns of 193, media errors 8. DRIVE HEALTH: 100% The temperature graph is nonsense because temp is not reported via a USB enclosure, but Rocket CP doesn't care, it's dutifully graphing a non-existent probe; that squiggle in the graph is a little UI flourish to let you know that if it could report activity it would appear actively. Do not try to comprehend the control panel's term "DISK" in context, except that the term "HEALTH" is about equally appropriate — the degrees of fakery and hype which attend Windows products is an endless source of amazement to me.

Sabrent support sagely observed that the drive probably went bad because of all these unsafe shutdowns. Cause or effect? You must decide...

SABRENT ACCESSORY HEAT-SINK

As I googled around for others' stories on NVMe failure, I was fed an Amazon promo for a Sabrent NVMe accessory heat-sink:

Sabrent Rocket Heat-sink.jpg


Maybe I needed one of these and just didn't know it?!

This awkward-looking device has piqued my curiosity because it makes the drive 20x more cumbersome to fit to a build and "Universal fit" in no way means that it fits universally.

The Sabrent accessory heat-sink has a clever design that connects the components on the bottom-side of the Sabrent drive to the cooler through the U-shaped mount. It also comes in designer colors and has cool-looking heat-pipes.

OTOH, the Asus slots provision a built-in thermal pad for the bottom of the drive. And the Samsung and Western Digital have no components on the bottom side of the drive!

Plus don't SSD NAND flash storage cells like it hot? Idk...

Then there is the matter of fit.

For example on this Asus z590 Hero:
  • M.2_1 slot position: Sabrent cooler will interfere with the CPU cooler, and this slot is already provisioned by Asus with an equally substantial heat-sink which fits under most CPU coolers.
  • M.2_2 slot is covered by the dGPU, so it can't go there.
  • M.2_3 slot could accept this cooler but this prohibits the use of the Asus-supplied heat sink for neighnor slo tM.2_4; a common cooler plate provisions both slots.
Wait-a-minute, all these PCIe gen4 boards already have a bottom support and thermal pad which will interfere with this accessory even if there is top clearance! And if you aren't using a gen 4 slot, why do you need this insane thermal solution? Again, I am very new to this NVMe tech, but a lot about the design feels like after-thoughts to a gamer-age of PC hardware which needs to end.
  • For the Samsung EVO 970 Plus and Western Digital SN750, the Asus-supplied heat-sinks seem completely sufficient.
One guy on web talked about how he had no problems with the fit by using a Dremel tool to grind away a portion of something to make this Sabrent cooler work. I just imagined metal dust and piles of fine metal shavings collecting around the slot connectors, then kept moving.

My regards are this Sabrent accessory cooler may appeal to some users the same way as do huge wheels on an SUV. But to me it's goofy enough that Sabrent's mere offering of this accessory betrays some other incomplete aspect of the industry's approach to thermal design.

Nonetheless I worried that maybe there's something I'm just not grokking? Maybe this drive's temp edge cases is serious stuff and if I'm not careful my data could go at any moment?!

This is where I become completely flummoxed. If such an accessory is required to protect the Rocket 4 from thermal failure, Sabrent is completely remiss at informing the customer of such requirement. While the Sabrent Rocket 4 packaging provides what one blogger reviewed as a 'sexy unboxing experience', the lavish packaging of the Rocket 4 includes no caution or even indication of the concern of thermal overload on data integrity. The user-guide does include a front-page note in black-and-white about optimal performance for non-PCIe 4.0 mainboards, but more as a matter of up-selling the customer on performance, rather than warning of failure:

Rocket-User's-Guide.png


My build is only running as PCIe gen 3 due to i9-10900K CPU, and as a gen 4 board, Asus supplies heat-sinks for all M.2 slots. as noted by the Sabrent guide.

So I don't think I missed anything.

So the Sabrent cooler cannot apply to my use-case no matter how universal the cooler otherwise is.

Regarding sexy packaging... By comparison with the lavish Sabrent, the Samsung and Western Digital drives arrive in minimal, but completely functional, clear plastic trays which probably require 1/10 the industrial energy input of the Sabrent's plastic-lined-metal-case-with-foam-tray-and-separate-paper-lined-die-cut-foam-cover-insert-plus finger-grip-tab-and-double-printed case decals. (Why does the slogan "Engineered in America" pop into mind at this moment?)

And you won't feel bad about throwing away a simple plastic tray as you hopefully you will never need it again because your drive will be in your PC helping you get work done, rather than in transit for a support team assessment as to whether your forever-losing your data deserves another—and maybe even a third—go-around.

ASUS Z590 MAXIMUS VIII HERO M.2 SLOT COOLERS

Asus-z590-Maximus-VIII-Hero-M.2-Coolers.png


ASIDE ON WESTERN DIGITAL GAMING MODE

I'm interested in how the industry thinks about looming power issues with NVMe drives.

The Western Digital SN750 dashboard (Windows only) has a "Gaming" setting which disables low-power mode. It's unclear as to various aspects of this feature:

Does it:

• Relax over-temp limiting?
• Run the device in high-power mode on laptops for desktop-like performance?
• Need to be reset at each boot?
• Give gamers a sense of magical ju-ju that tech-wizards are blowing the doors off the load-times for cut-scenes?!

Reviews of the feature suggest no benefit in typical desktop scenarios and wonderment about the purpose of the feature. based on my quick tests, it's academic because the drive easily reaches full PCIe gen 3 bus limit, and stays reassuringly cool doing so.

The term "Gaming" seems to be a euphemistic market designator for customers who are expected out of enthusiasm and ignorance to tolerate a lot vapid pandering and failure, so I've decided to stay away from any such features.

CONCLUSION FOR MY BUILD

I have been through hell with the Rocket 4.

IMG_2299+.jpg
 
Last edited:
Did anyone with Z590 and Comet Lake (not Rocket Lake) ever get their iGPU fully working and accelerated?

I have Rocket Lake so I can't test that, but if it works I would like to add support for it in my HackinDROM Z590 Vision D for Comet Lake users.
 
Back
Top